Sorry, late to the party, but wanted to pass along some info that may be helpful to folks. This is something that I have experimented with in years past and why I have always been verbal that I am not a fan of any truncation below 80 ms for close mic'd files, and a minimum of 200-300 ms for ambient files. Hopefully the below will calm things down a bit and give some perspective from my vantage point. Additionally, I have yet to even remotely dabble in programming/coding, so my view point and knowledge is purely based upon observation of experimentation and use with existing technologies that I make these IR's for, but do not have any hand in at a programming level. I cannot speak intelligently on why this is the way it is, just that it is the way it is.
This phenomenon is not hokum, and can be demonstrated using standard convolution loaders from many years past. Shown below is my D120F IR represented earlier in this thread, loaded with Voxengo's Perfect Space that Aleksey wrote for Cakewalk (and I assume is a stripped down derivative of Pristine Space). As shown, I printed one response at 100 ms, then cut it to 20 ms:
That plugin allows you to truncate the tail at will to varying degrees, and as you pull the value down you can watch the plot transform in varying manners. To keep things clean and simple, I just replicated the values that Cliff posted earlier.
This phenomenon is known to others, including myself, and has been for a few years now. I, AnalogInTheBox, and DMG Audio have even been using longer tail's in such things as hardware sampling and replicating (EQ's, mic preamps, etc) VST plugins, let alone guitar cabinet IR's, to increase accuracy in the extreme ends of the spectrum (high and low). In the EQ realm, it is known that the lower the sample length of the IR, the less high frequency information will be present as well. In Nebula, those of us "in the know" typically truncate the fundamental kernel to 100 ms instead of the stock 50 ms as, again in my experience, below 80 ms things start getting less accurate. The owner of DMG has very adamantly pointed this out in pro audio forums upon the release of his EQuilibrium plugin, and has a "quality" setting that in effect raises or lowers the amount of samples in the impulse response which also raises or lowers how much resource overhead the plugin consumes.
This phenomenon is already available to be had if you are a recordist and use DAW based plugins. This is why none of my cab IR's are ever, nor have they ever been below 100 ms.
This phenomenon is not present on all files, and from recent observation seems to be vastly dependent on the internal resonance of the cabinet. Upon quick inspection, I do not see this happening, at least not in this amount, on the open back 2x12 IR's I am creating right now. Maybe this is why Jay got fired up on TGP - the files he is using for demonstration do not exhibit this, and I can vouch for that side as well because not all files are as extreme as in the examples shown. This example shows the difference between the new 500 ms ("UltraRes") and the 20 ms castration with a Jensen P12R in a the aforementioned open back combo:
I've only sampled about a half dozen speakers for this cab so far, but all seem to have a much smaller decrease in low frequencies as the sample length goes down.
Higher tail lengths increase the information in the higher frequencies as well, in my experience. When I spoke with Cliff on this matter, I tried 100 ms vs 500 ms and in the DAW noted that I heard an increase in the hypersonic range. That was without looking at anything and just listening. Upon glancing at a graph in the last few days, this was verified with an increase in volume of frequencies above 12 kHz. This will mean a small increase in the "air" band in some cases, at least has been my experience in the DAW realm.
MOST importantly, in MY opinion, this feature will finally add the room details and information for those that wish to load IR's in these magic black boxes. This is make or break from a capture standpoint - IR makers with bad rooms won't be able to hide from them and will suffer even worse sounding IR's, however those of us who have meticulously tuned our rooms and sources within our rooms will have users be able to benefit from that without having to have a computer in tow. To this end, those who may not want this I assume can always just bump things down to HighRes to take the room back out again.
All of this argument and elitism over this, in my opinion, is completely ridiculous and based on observations from incomplete information. To this end, it would not be anyone's fault to fall heavily on one side of the coin or the other as the data they have in front of them would prove their side of the debate. I have the luxury of being able to sit on both sides peacefully as I feel I have data that shows how each side can have a valid point, and despite my busy schedule and need to rest up for this upcoming weekend, wanted to post all this in hopes to bring the conversation to more positive directions.
I'm very happy to see the resolution of the cab block go up to where I can happily recommend usage, and this puts the FAS platform even further ahead of the pack so far as I am concerned. Even my fellow tube amp user friends of mine won't hesitate to let you know that I have warned them many times not to use my IR's in their rackmount load box, and to just send the line out to their interface and use full length IR's, citing these exact principals as my reason why.
Hope this helps, if not, take your weapon, strike this post down with all of your hatred and your journey towards the dark side will be complete! :lol
Peace!